Learning system design from ants

How ARE we to think of healthcare systems? It has been fashionable to think of them supertankers — the images conjured up of something big, slow and as the politicians and managers were wont to say, slow to turn, so (to cite Piet Hein’s TTT) things take time. But such thinking is wrong-headed and always has been — it reflected a top-down technocratic mind-set that saw healthcare as rational and plannable.  I have never bought this argument, and argued that healthcare systems should be conceptualised as a school of fish, as they can turn on a dime! We have much to learn from a swam of bees, a colony of ants or a flock of birds. (and even a school of fish). The bigger and more complex a system becomes (like healthcare systems) the least likely it is to be amenable to notions of levers to push or string to pull type thinking — but such thinking veritably oozes from academe where reductionist and linear analytical models are easier to research, but fail to take account of reality itself.  (the academic group that seems to consistently get things wrong for similar reasons is economics, and surprise, surprise, many people think that health economics has something to tell us about health reform — but similar models and thinking are pervasive!)

Now, Peter Miller’s book, Smart Swarm has received laudable coverage in the Economist newspaper. The book and many articles in the popular press have highlighted the efficient design that comes from the apparently unplanned but linked behaviour of individual generally unintelligent ants or bees, which collectively bring order to complex natural environments.

For our cherished health system planners, it suggests that they have overstated their impact and relevance, if indeed they ever were really effective. For me, it suggests that health policy has become a bit too much like the old Soviet bread planners, thinking that since people needed the bread, planning for it would ensure is got to people’s table, but of course the planning was part of the problem, not the solution. (I’m trying not to make the bread an issue of markets but of coordinated behaviour of linked systems.)

The best way to understand complex systems is to embed intelligence within the behaviour of the bits that make it up, rather than impose it from above, or fruitlessly planned in. The key factor which makes these distributed systems work is the ability to exchange information — planners create funnels through which information flows and if you get the funnels wrong, the system fails to optimise, or indeed work at all! By allowing parts to exchange information easily, on an as-needed basis and act accordingly, coordinated behaviours emerge, which effectively bring the desired order without some remote planner deciding how it should work. It all comes down to information flow and exchange.

And so to health systems.  We all want joined up, linked, coordinated healthcare; that patients seen in clinic A who go to clinic B can be seen by people who have information about you; that when you show up for your operation, they know you’re coming, and so on. Healthcare systems are really all about patients, but we plan then on the basis of the behaviour of health professionals, who actually communicate with each other quite a lot — and indeed, construct informal systems to make the healthcare system work better often despite formal planned structures . One might say they behave like the bees and ants by simply getting on with things. It does raise the question of what roles are needed within healthcare systems to ensure the flow of information — this is usually seen as a reason for managers, but managerial models frequently fail to understand the purposeful behaviour of interconnecting systems as management is a reductionist organisational notion.

The information that patients carry is critical, but generally not accessed — it is important to realise that only the patient has experience of the whole care pathway, not the health professionals within it. This is important information that is lost within formally planned systems, which focus on structuring care, rather than the flow of information that links the bits together. I call this information structure the “healthcare information value chain” and it is the most important, but generally least understood, aspect of healthcare systems, as embedded within this value chain is the information needed for the various components of a healthcare system to coordinate their activities, without the great planner in the sky. This information value chain is not some construction of an IT system, it is information used in the course of people going about their work — of the ant hill at work. It is worth noting that ants or bees manage complex systems without IT systems, but researchers have needed IT systems to understand them!

Want to know more?

Ants may have an edge when designing efficient systems, a commentary on US health system reform, by  Joseph Reisert.

Emergence: The connected lives of ants, brains, cities, and software, a book by Steven Johnson.

It may also be helpful to understand how complex adaptive systems work: for instance,

The US National Academy of Medicine has thought about healthcare as a complex adaptive system in “Health Care as a Complex Adaptive System: Implications for Design and Management”. [here]

Implementation Science is an open source journal that has articles on this subject, including this example [here] on making change in healthcare settings.

Regretfully, much useful literature is not available to the informed or interested public, as it is squirrelled away in the academic journals for which the publishers require passwords, subscriptions etc. in order to access.  Any research that has been funded from public sources should be available for public access in open source locations or journals. I will not cite reference material that is not generally available to the public. If authors have material on this subject that they would like to enable public access to, please send me the links to be added to this (very) selective list.

Something NICE needs to do

NEWS FLASH: Setting a minimum price for a unit of alcohol would help tackle Britain’s drink problem, health advisers are expected to recommend. The National Institute for Health and Clinical Excellence (Nice) will include the advice in its guidance on how to crack down on problem drinking. (1 June 2010)

This commentary is not on whether to set a minimum price for alcohol. This is a comment about expansion of the scope of NICE’s mandate.

What is NICE for and why are they now becoming involved in more fundamental health policy matters? Under the rubric of health excellence, one assumes they are pushing this as far as they can possibly go.

NICE is really a disguised authoritarian advisory body because of their lack of proper public accountability coupled with their privileged access to ministers in government.

NICE are not ‘health advisors’; they are a fourth hurdle advisory body with a focus on what works in healthcare service delivery, such as medicines and device technologies. By moving outside this, they are creating the impression that any area of health interest can be subjected to their methodologies. Indeed, that all matters of policy can be reduced to a QALY analysis and some economic modelling. No doubt at some point, they will pass judgement on the health impact of the national speed limit,  the salt content of food, the pub opening hours, as long as there is some way to tie the analysis to a health outcome. Invoking their brand of technocratic thinking to replace the fine art of public consultation is hardly the way ahead — that there is some evidence for the benefits or costs, does not lead inexorably to the conclusion that health policy should change.  Running health policy by the numbers in this way guts the democratic process for deciding social priorities.

This all-purpose extension of the mandate of NICE is not a good thing, for democracy or for health policy in the UK.

Lawmakers burning discredited health policies

‘Significantly, the core principles (and expected savings) of polysystems have proven difficult to achieve with more focus on the buildings rather than the changes to care and behaviours.’

So says a recently hitherto secret report that NHS London (UK) has been sitting on. What a surprise though. Healthcare change is difficult and the focus on so-called polysystems missed the point. In origin, they are really polyclinics, and well-designed would cut admission rates to secondary care; they would also bulk up on specialist services, including day-care work and short-stay facilities.

A bureaucratic orientation driven by doctrinaire thinking and misaligned incentives are clearly to blame, plus, of course, a fear, within the NHS of actual service reconfiguration and change that alters the structure and nature of clinical work.

Whether the new UK government coalition should actually stop the polysystems (a euphemism too far, I fear) is another question, as the underlying logic, used successfully in other countries is sound.  What really failed was management, and the vaunted commissioning system, which failed to demand, perhaps even conceptualise, service changes. No doubt, resistance from the clinical professions may have no small part in failure, but clinicians are been substantially disenfranchised from NHS reform, with the top-down, initiative driven thinking.

Less is more. Few but more substantial changes, may ultimately lead to the service and quality improvements.

Polyclinics are a missed opportunity, and having been badly conceived are now a tainted option. The political pull back to the status quo becomes a real a risk, when in fact greater effort than ever is needed to improve service delivery and productivity.

Beware digital errors as they can bite

We all know accidents (unusual occurances in healthcare) can happen. Where systems are involved, errors can arise from how a system works, the way the various bits mesh, the knowledge and training of everyone involved working together.  It is no real surprise that some errors arise from the technologies that we use. In particular, health information technology systems can cause new types of errors and mistakes, beyond just not working properly.

In the US, the Health IT Policy Committee has proposed establishing a database to track potential safety risks related to IT systems.  These risks include:

• hardware and software failure and bugs
• workflow interactions between staff and users
• interoperability problems
• implementation and training deficits.

Since healthcare work is complex, the workflow risks are particularly complex and can arise from, for instance, inaccurately understanding how a manual system achieves its results, and thereby designing a software-based system that fails to do just that. There is a funny little thing that happens when a patient sees a doctor; the doctor often will use writing a prescription to terminate the patient encounter — tearing the piece of paper off the tab, a swirl of signature and handing the slip to the patient leads to the patient leaving, a neat way to end the consultation.

In an automated system (electronic prescribing, for instance), the consultation is not terminated in this behavioural manner, but involves essentially hitting the return key on the keyboard to enter the required prescription data in the system, and perhaps handing (or not) the patient a copy — but the Rx is off on electronic wings to the pharmacy for dispensing. There is an error that can occur if the doctor does not hit the return key between patients — the Rx list builds up, from patient to patient, until the return key gets hit (unless some sort of failsafe has been built in); this error actually happened and it was an alert pharmacist commenting to the patient that the doctor had added a lot of new drugs that the alarm was raised. Perhaps the patient should have been more distrustful, too.

We must be mindful of risk and error in any kind of technology, but particularly in systems where it is very hard to look inside the black box of software code.

I wrote a paper on digital risk some years ago, which can be found here: Patient Safety and Digital Risk. I have also raised the issue of risk in the even blacker box of predictive algorithms used to data mine record systems and profile risk of patients and this can be found here: Predictive Health. This second paper suggested that software may need to be subjected to comparable regulatory review like a medical device.

Just because you can’t drop it on your foot, doesn’t mean something can’t be dangerous.